Author image Mark Overmeer
and 1 contributors

NAME

Log::Report::Template - template toolkit with translations

INHERITANCE

 Log::Report::Template
   is a Template

SYNOPSIS

  use Log::Report::Template;
  my $templater = Log::Report::Template->new(%config);
  $templater->addTextdomain(name => "Tic", lexicon => ...);
  $templater->process('template_file.tt', \%vars);

DESCRIPTION

This module extends Template, which is the core of Template Toolkit. The main addition is support for translations via the translation framework offered by Log::Report.

You add translations to a template system, by adding calls to some translation function (by default called 'loc()') to your template text. That function will perform dark magic to collect the translation from translation tables, and fill in values. For instance:

  <div>Price: [% price %]</div>          # no translation
  <div>[% loc("Price: {price}") %]</div> # translation optional

It's quite a lot of work to make your templates translatable. Please read the "DETAILS" section before you start using this module.

METHODS

Constructors

Log::Report::Template->new(%options)

Create a new translator object. You may pass the options as HASH or PAIRS. By convension, all Template Toolkit options are in capitals. Read Template::Config about what they mean. Extension options are all in lower-case.

In a web-environment, you want to start this before your webserver starts forking.

 -Option           --Default
  modifiers          []
  processing_errors  'NATIVE'
  template_syntax    'HTML'
modifiers => ARRAY

Add a list of modifiers to the default set. Modifiers are part of the formatting process, when values get inserted in the translated string. Read "Formatter value modifiers".

processing_errors => 'NATIVE'|'EXCEPTION'

The Template toolkit infrastructure handles errors carefully: undef is returned and you need to call error() to collect it.

template_syntax => 'UNKNOWN'|'HTML'

Linked to String::Print::new(encode_for): the output of the translation is HTML encoded. Read "Translation into HTML"

Attributes

$obj->formatter()

Get the String::Print object which formats the messages.

Handling text domains

$obj->addTextdomain(%options)

Create a new Log::Report::Template::Textdomain object. See its new() method for the options.

Additional facts about the options: you may specify only_in_directory as a path. Those directories must be in the INCLUDE_PATH as well. The (domain) name must be unique, and the function not yet in use.

example:

  my $domain = $templater->addTextdomain(
    name     => 'my-project',
    function => 'loc',   # default
    lexicon  => $dir,    # location of translation tables
  );
$obj->extract(%options)

Extract message ids from the templates, and register them to the lexicon. Read section "Extracting PO-files" how to use this method.

 -Option        --Default
  charset         'UTF-8'
  filename_match  qr/\.tt2?$/
  filenames       undef
  show_stats      <false>
  write_tables    <true>
charset => CHARSET
filename_match => RegEx

Process all files from the INCLUDE_PATH directories which match this regular expression.

filenames => FILENAME|ARRAY

By default, all filenames from the INCLUDE_PATH directories which match the filename_match are processed, but you may explicitly create a subset by hand.

show_stats => BOOLEAN

Show statistics about the processing of the template files.

write_tables => BOOLEAN

When false, the po-files will not get updated.

Template filters

Some common activities in templates are harder when translation is needed. A few TT filters are provided to easy the process.

Filter: cols

A typical example of an HTML component which needs translation is

  <tr><td>Price:</td><td>20 £</td></tr>

Both the price text as value need to be translated. In plain perl (with Log::Report) you would write

  __x"Price: {price £}", price => $product->price    # or
  __x"Price: {p.price £}", p => $product;

In HTML, there seems to be the need for two separate translations, may in the program code. This module (actually String::Print) can be trained to convert money during translation, because '£' is a modifier. The translation for Dutch (via a PO table) could be

   "Prijs: {p.price €}"

SO: we want to get both table fields in one translation. Try this:

  <tr>[% loc("Price:\t{p.price £}" | cols %]</tr>

In the translation table, you have to place the tabs (backslash-t) as well.

There are two main forms of cols. The first form is the containerizer: pass 'cols' a list of container names. The fields in the input string (as separated by tabs) are wrapped in the named container. The last container name will be reused for all remaining columns. By default, everything is wrapped in 'td' containers.

  "a\tb\tc" | cols             <td>a</td><td>b</td><td>c</td>
  "a\tb\tc" | cols('td')       same
  "a\tb\tc" | cols('th', 'td') <th>a</th><td>b</td><td>c</td>
  "a"       | cols('div')      <div>a</div>
  loc("a")  | cols('div')      <div>xxxx</div>

The second form has one pattern, which contains (at least one) '$1' replacement positions. Missing columns for positional parameters will be left blank.

  "a\tb\tc" | cols('#$3#$1#')  #c#a#
  "a"       | cols('#$3#$1#')  ##a#
  loc("a")  | cols('#$3#$1#')  #mies#aap#
Filter: br

Some translations will produce more than one line of text. Add '<br>' after each of them.

  [% loc('intro-text') | br %]
  [% | br %][% intro_text %][% END %]

Formatter value modifiers

Modifiers simplify the display of values. Read the section about modifiers in String::Print. Here, only some examples are shown.

You can achieve the same transformation with TT vmethods, or with the perl code which drives your website. The advantange is that you can translate them. And they are quite readible.

POSIX format %-10s, %2.4f, etc

Exactly like format of the perl's internal printf() (which is actually being called to do the formatting)

Examples:

 # pi in two decimals
 [% loc("π = {pi %.2f}", pi => 3.14157) %]

 # show int, no fraction. filesize is a template variable
 [% loc("file size {size %d}", size => filesize + 0.5) %]
BYTES

Convert a file size into a nice human readible format.

Examples:

  # filesize and fn are passed as variables to the templater
  [% loc("downloaded {size BYTES} {fn}\n", size => fs, fn => fn) %]
  # may produce:   "  0 B", "25 MB", "1.5 GB", etc
Time-formatting YEAR, DATE, TIME, DT

Accept various time syntaxes as value, and translate them into standard formats: year only, date in YYYY-MM-DD, time as 'HH::MM::SS', and various DateTime formats:

Examples:

  # shows 'Copyright 2017'
  [% loc("Copyright {today YEAR}", today => '2017-06-26') %]
 
  # shows 'Created: 2017-06-26'
  [% loc("Created: {now DATE}", now => '2017-06-26 00:24:15') %]
  
  # shows 'Night: 00:24:15'
  [% loc("Night: {now TIME}", now => '2017-06-26 00:24:15') %]
  
  # shows 'Mon Jun 26 00:28:50 CEST 2017'
  [% loc("Stamp: {now DT(ASC)}", now => 1498429696) %]
Default //"string", //'string', or //word

When a parameter has no value or is an empty string, the word or string will take its place.

  [% loc("visitors: {count //0}", count => 3) %]
  [% loc("published: {date DT//'not yet'}", date => '') %]
  [% loc("copyright: {year//2017 YEAR}", year => '2018') %]
  [% loc("price: {price//5 EUR}", price => product.price %]
  [% loc("price: {price EUR//unknown}", price => 3 %]

Template (Toolkit) base-class

The details of the following functions can be found in the Template manual page. They are included here for reference only.

$obj->error()

If the 'processing_errors' option is 'NATIVE' (default), you have to collect the error like this:

 $tt->process($template_fn, $vars, ...)
    or die $tt->error;

When the 'procesing_errors' option is set to 'EXCEPTION', the error is translated into a Log::Report::Exception:

  use Log::Report;
  try { $tt->process($template_fn, $vars, ...) };
  print $@->wasFatal if $@;

In the latter solution, the try() is probably only on the level of the highest level: the request handler which catches all kinds of serious errors at once.

$obj->process( $template, [\%vars, $output, \%options] )

Process the $template into $output, filling in the %vars.

DETAILS

Textdomains

This module uses standard gettext PO-translation tables via the Log::Report::Lexicon distribution. An important role here is for the 'textdomain': the name of the set of translation tables.

For code, you say "use Log::Report '<textdomain>;" in each related module (pm file). We cannot do achieve comparible syntax with Template Toolkit: you must specify the textdomain before the templates get processed.

Your website may contain multiple separate sets of templates. For instance, a standard website implementation with some local extensions. The only way to get that to work, is by using different translation functions: one textdomain may use 'loc()', where an other uses 'L()'.

Supported syntax

Translation syntax

Let say that your translation function is called 'loc', which is the default name. Then, you can use that name as simple function:

  [% loc("msgid", key => value, ...) %]
  [% loc('msgid', key => value, ...) %]
  [% loc("msgid|plural", count, key => value, ...) %]
 
  [% INCLUDE
       title = loc('something')
   %]

But also as filter. Although filters and functions work differently internally in Template Toolkit, it is convenient to permit both syntaxes.

  [% | loc(key => value, ...) %]msgid[% END %]
  [% 'msgid' | loc(key => value) %]
  [% "msgid" | loc(key => value) %]

As examples

  [% loc("hi {n}", n => name) %]
  [% | loc(n => name) %]hi {n}[% END %]
  [% "hi {n}" | loc(n => name) %]

These syntaxes work exacly like translations with Log::Report for your Perl programs. Compare this with:

  __x"hi {n}", n => name;    # equivalent to
  __x("hi {n}", n => name);  # replace __x() by loc()

Translation syntax, more magic

With TT, we can add a simplificition which we cannot offer for Perl translations: TT variables are dynamic and stored in the stash which we can access. Therefore, we can lookup "accidentally" missed parameters.

  [% SET name = 'John Doe' %]
  [% loc("Hi {name}", name => name) %]  # looks silly
  [% loc("Hi {name}") %]                # uses TT stash directly

Sometimes, computation of objects is expensive: you never know. So, you may try to avoid repeated computation. In the follow example, "soldOn" is collected/computed twice:

  [% IF product.soldOn %]
  <td>[% loc("Sold on {product.soldOn DATE}")</td>
  [% END %]

The performance is predictable optimal with:

  [% sold_on = product.soldOn; IF sold_on %]
  <td>[% loc("Sold on {sold_on DATE}")</td>
  [% END %]

Translation into HTML

Usually, when data is passed from the program's internal to the template, it should get encoded into HTML to escape some characters. Typical TT code:

  Title&gt; [% title | html %]

When your insert is produced by the localizer, you can do this as well (set template_syntax to 'UNKNOWN' first)

  [% loc("Title> {t}", t => title) | html %]

The default TT syntax is 'HTML', which will circumvent the need to use the html filter. In that default case, you only say:

  [% loc("Title> {t}", t => title) %]
  [% loc("Title> {title}") %]  # short form, see previous section

When the title is already escaped for HTML, you can circumvent that by using tags which end on 'html':

  [% loc("Title> {t_html}", t_html => title) %]

  [% SET title_html = html(title) %]
  [% loc("Title> {title_html}") %]

Extracting PO-files

You may define a textdomain without doing any translations (yet) However, when you start translating, you will need to maintain translation tables which are in PO-format. PO-files can be maintained with a wide variety of tools, for instance poedit, Pootle, virtaal, GTranslator, Lokalize, or Webtranslateit.

Setting-up translations

Start with desiging a domain structure. Probably, you want to create a separate domain for the templates (external texts in many languages) and your Perl program (internal texts with few languages).

Pick a lexicon directory, which is also inside your version control setup, for instance your GIT repository. Some po-editors can work together with various version control systems.

Now, start using this module. There are two ways: either by creating it as object, or by extension.

  ### As object
  # Somewhere in your code
  use Log::Report::Template;
  my $templater = Log::Report::Template->new(%config);
  $templater->addTextdomain(...);

  $templater->process('template_file.tt', \%vars); # runtime
  $templater->extract(...);    # rarely, "off-line"

Some way or another, you want to be able to share the creation of the templater and configuration of the textdomain between the run-time use and the irregular (off-line) extraction of msgids.

The alternative is via extension:

  ### By extension
  # Somewhere in your code:
  use My::Template;
  my $templater = My::Template->new;
  $templater->process('template_file.tt', \%vars);
  
  # File lib/My/Template.pm
  package My::Template;
  use parent 'Log::Report::Template';

  sub init($) {
     my ($self, $args) = @_;
     # add %config into %$args
     $self->SUPER::init($args);
     $self->addTextdomain(...);
     $self;
  }

  1;

The second solution requires a little bit of experience with OO, but is easier to maintain and to share.

adding a new language

The first time you run extract(), you will see a file being created in $lexicon/$textdomain-$charset.po. That file will be left empty: copy it to start a new translation.

There are many ways to structure PO-files. Which structure used, is detected automatically by Log::Report::Lexicon. My personal preference is $lexicon/$textdomain/$language-$charset.po. On Unix-like systems, you would do:

  # Start a new language
  mkdir mylexicon/mydomain
  cp mylexicon/mydomain-utf8.po mylexicon/mydomain/nl_NL-utf8.po 
  
  # fill the nl_NL-utf8.po file with the translation
  poedit mylexicon/mydomain/nl_NL-utf8.po
  
  # add the file to your version control system
  git add mylexicon/mydomain/nl_NL-utf8.po
  

Now, when your program sets the locale to 'nl-NL', it should start translating to Dutch. If it doesn't, it is not always easy to figure-out what is wrong...

Keeping translations up to date

You have to call extract() when msgids have changed or added, to have the PO-tables updated. The language specific tables will get updated automatically... look for msgids which are 'fuzzy' (need update)

You may also use the external program xgettext-perl, which is shipped with the Log::Report::Lexicon distribution.

More performance via MO-files

PO-files are quite large. You can reduce the translation table size by creating a binary "MO"-file for each of them. Log::Report::Lexicon will prefer mo files, if it encounters them, but generation is not (yet) organized via Log::Report components. Search for "msgfmt" as separate tool or CPAN module.

SEE ALSO

This module is part of Log-Report-Template distribution version 0.13, built on January 23, 2018. Website: http://perl.overmeer.net/CPAN/

LICENSE

Copyrights 2017-2018 by [Mark Overmeer]. For other contributors see ChangeLog.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself. See http://dev.perl.org/licenses/